An AI-Based Intrusion Prevention System to Enhance Cloud Security
P Sarumathy, S. Alamelu Alias Rajasree, A. Chandrasekar · 2025
Data breaches have become increasingly prevalent in cloud environments, demanding robust security measures. In an era where data breaches pose significant threats to organizational security and individual privacy, the protection of sensitive data has become paramount. A comprehensive strategy is proposed for analyzing sensitive data and preventing intrusions using a multi-modal framework. The methodology integrates AI techniques, Long Short-Term Memory (LSTM) networks for classification of data as sensitive/non-sensitive and Parameter Efficent Fine-Tuning(PEFT) for categorization of sensitive data. Additionally, organization-specific regular expression (regex) pattern matching algorithms are utilized. Together, these components form an Intrusion Prevention System (IPS) capable of quickly identifying and alerting against unauthorized changes, with a focus on detecting sensitive data. The system also addresses pivot attacks, where attackers exploit vulnerabilities to access data across servers within the same network. By representing system components and their relationships as graphs, our approach enhances the detection and prevention of such attacks. This graphical representation facilitates a better understanding of component relationships and helps identify redundant storage elements, optimizing resource allocation. Our rigorous evaluation demonstrates the effectiveness of our approach in mitigating data breach risks and protecting sensitive information.